Discuss as:

How does the NSA's phone-record program work?

Here is how we THINK this program works, based on several calls.

The National Security Agency, with approval from the Foreign Intelligence Surveillance Act court, goes to the phone companies and says: Every day, pump your data about phone calls into our big government tank -- only phone numbers (not names), along with other data about the calls, such as where they came from, how long they lasted, what numbers were dialed, and so on.

The government does this for two reasons. First, the phone companies themselves retain this data only for 30-90 days. Second, having this data aggregated in one place avoids the need, when checking out a specific number, to have to go get a court order for each phone company to track a specific number, when the government does not know what company handled that number.

(USA Today reported on this program back in 2006.)

So the order made public by the Guardian covers part one of this process. 

Part two, however, involves the authority to dive into the government's big tank of phone data to check out a specific number. It does this under protocols and internal controls that are approved in advance and must be reviewed by the FISA court periodically. This means that the tank can be accessed only for specific reasons -- not to generally roam through.

So, for example, when the police in London bust up a terror cell and find a U.S. number was on one of their cell phones, the U.S. government can then check that number against its big tank of phone data, without having to go to every carrier.

But there are several questions for the administration to answer, among them:

-- How does the FOREIGN intelligence surveillance court have authority to order the phone companies to turn over records of DOMESTIC calls, as well as international ones?

-- How does it satisfy the Fourth Amendment to require a company to turn over all its data in advance?

-- And what's the Fourth Amendment authority to tell a court what specific searches were conducted on that data after the fact, instead of in advance?